{"id":"8fbe6cd8-21c7-4e40-bd11-8a643b0bd1c1","arxiv_id":"2411.18965","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Dynamic extensions that homogenize transport velocities enable backstepping controllers for hyperbolic PDEs to assign general closed-loop dynamics and achieve complete input-output decoupling.","lead":"The paper introduces dynamic extensions, extra transport equations placed inside the controller, so that backstepping control of hyperbolic PDE systems can assign arbitrary closed-loop dynamics and fully decouple inputs from outputs. This gives control engineers a systematic new tool that static feedback cannot provide, with extensions to PDE-ODE systems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The decoupling construction in §4.2 rests on an unproven existence claim for the reduced kernel equations; the general Lemma 4 does not cover this subset, and the stated conditions do not explicitly rule out a missing \\bar A^+_0 contribution.","rationale":"The paper's overall architecture is coherent: a dynamic extension homogenizes transport velocities, after which a static feedback of the extended state can assign general closed-loop dynamics, and the decoupling construction is the most novel part of the central claim. The reader identified the weakest point as the unproven well-posedness of the reduced kernel equations in Section 4.2, and I agree that this is the load-bearing gap. My stress test sharpens the concern: the reduced kernel problem is not covered by Lemma 4, and the fact that \\chi^+ is left unchanged introduces a potential extra term involving \\bar A^+_0 unless L^{-+}=0. This is likely recoverable by a direct characteristics argument, but the paper does not supply it. Since the missing step is a proof gap rather than a demonstrated counterexample, the appropriate disposition remains conditional: the main claims are plausible and largely supported, but the decoupling result should not be fully accepted until the reduced kernel equations are shown to admit a solution and the derivation of (43) is verified. This does not change the reader's verdict, so no adjustment is proposed.","tokens_in":15956,"tokens_out":37621,"duration_ms":340288,"concrete_test":"Independently derive the kernel equations for (42) by substituting the transformation into (28) and collecting the coefficients of \\bar\\chi^-, \\chi^+, and \\bar\\chi^-(0). Check whether the resulting equations are exactly (38a), (38b), (38d) with \\bar B=0, or whether an extra \\int_0^z L^{-+}\\bar A^+_0 d\\zeta term appears. For the Section 5 example, solve the reduced characteristic equations for L^{--} with L^{-+}=0 and verify by simulation that y(t)=\\bar v(t-\\phi^-_{n_-}(1)) holds in the closed loop with the feedback (43d). If the derived equations differ from (38a)-(38d), or if the simulation fails, the decoupling claim as stated requires revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central decoupling claim is made in §4.2: if L^{--} and L^{-+} solve (38a), (38b), (38d) with \\bar B=0, then (42) maps (28) into (43) and yields y(t)=\\bar v(t-\\phi^-_{n_-}(1)). No theorem in the paper proves that this reduced set of kernel equations has a solution. Lemma 4 addresses the full system (38) with the artificial BC (40), not the reduced subset used here. The gap is substantive: because \\chi^+ is left unchanged, the target \\chi^+ dynamics retains \\bar A^+_0(z)\\bar\\chi^-(0,t). Substitution of (42) into (28) therefore produces an extra term \\int_0^z L^{-+}(z,\\zeta)\\bar A^+_0(\\zeta)d\\zeta\\,\\bar\\chi^-(0,t) unless L^{-+}=0. The paper does not show that (38a)-(38d) force L^{-+}=0, although this is plausible: (38b) prescribes L^{-+}(z,z)=0, and the characteristics of (38a) for \\bar B=0 run from the diagonal to \\zeta=0. Without either proving that degeneracy or including the extra term, the derivation of the decoupled target dynamics is incomplete.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a dynamic-extension approach for backstepping boundary control of general heterodirectional hyperbolic PDE systems. A preliminary Volterra transformation removes the in-domain coupling, and a transport-type dynamic extension then homogenizes the transport velocities so that all negative-direction components share one velocity and all positive-direction components share another. The resulting extended system (28) is used to design a static feedback of the extended state, which assigns a target system with arbitrary in-domain couplings (Section 4.1) and, in Section 4.2, to achieve complete input-output decoupling. A sketch for hyperbolic PDE-ODE systems is given in Section 4.3, and a numerical example illustrates the decoupling design.","tokens_in":16282,"tokens_out":17161,"duration_ms":152030,"significance":"If correct, the paper makes a useful methodological contribution: it transfers the finite-dimensional concept of dynamic extensions to the backstepping framework for hyperbolic PDEs and shows that this enlarges the achievable closed-loop dynamics beyond static state feedback. The construction of the dynamic extension is explicit and modular, and the stability argument in Section 4.1 follows standard backstepping reasoning with kernel well-posedness cited from [11]. The paper also gives a concrete numerical demonstration of the decoupling idea. However, the central decoupling result in Section 4.2 rests on an unproved well-posedness claim for a reduced kernel system and on an omitted cancellation that is needed for the claimed decoupled form (43); these issues must be fixed before the main contribution can be fully accepted. The PDE-ODE transfer in Section 4.3 is only sketched rather than proved.","major_comments":[{"comment":"The decoupling construction is incomplete as written. Substituting the transformation (42) into the positive-component equation of (28a) introduces the term \\int_0^z L^{-+}(z,\\zeta)\\bar A^+_0(\\zeta)\\,d\\zeta\\,\\bar\\chi^-(0,t) into the equation for \\bar\\chi^-, because \\chi^+ obeys the PDE with source \\bar A^+_0(z)\\chi^-(0,t). This term is absent from the claimed target (43a) unless L^{-+}\\equiv0. The paper does not prove that the reduced kernel equations (38a),(38b),(38d) with \\bar B(z)=0 force L^{-+}=0, nor that the remaining reduced problem for L^{--} is well-posed. Lemma 4 addresses the full system (38) with the artificial boundary condition (40), not the reduced subset used here. The claim that (42) maps (28) into (43) therefore needs a dedicated lemma establishing L^{-+}=0 and the well-posedness of the reduced kernel equations; alternatively, if L^{-+} is not zero, the target dynamics must be revised.","section":"Section 4.2, Eqs. (42)-(43)"},{"comment":"The abstract states that the modularity of the design allows a straightforward transfer of all results to hyperbolic PDE-ODE systems, but Section 4.3 provides no theorem, no kernel equations for the combined plant-ODE problem, and no proof of stability or decoupling for the PDE-ODE case. The text only cites [7] and states that the extension is 'mostly straightforward.' If the PDE-ODE transfer is intended as a contribution, precise statements and proofs should be given; otherwise the section should be explicitly labeled as a sketch or outlook.","section":"Section 4.3"},{"comment":"The proof of Lemma 4 is only a citation to [11] together with a short sketch. Since the homogenized system has repeated velocities inside each block of \\bar\\Lambda, the characteristic structure degenerates relative to the distinct-velocity case in [11], so the well-posedness of (38) should be verified explicitly for this repeated-velocity setting, including the role of the artificial boundary condition (40). The sketch in Figure 2 is plausible, but the paper should state the precise regularity and uniqueness result it claims.","section":"Section 4.1, Lemma 4"}],"minor_comments":[{"comment":"Footnote 1 says that the homogenized kernel equations are well-posed without additional artificial boundary conditions, while Lemma 4 explicitly introduces the artificial boundary condition (40). This apparent contradiction should be resolved by clarifying exactly which boundary conditions are needed.","section":"Section 4.1, Lemma 4 and footnote 1"},{"comment":"The claim that the dynamic feedback does not increase the minimum control time T_min is stated without proof. While the target system (34) with zero design matrices is clearly finite-time stable in time T_min, the statement that the original state x(z,t) vanishes for t>T_min requires tracking the effects of the preliminary transformation (5) and the dynamic extension; please add a proof or a precise reference.","section":"Remark 7"},{"comment":"The notion 'asymptotically stable pointwise in space' is used in the statement of Theorem 6 but is not defined, and no constructive conditions on B0, B1, B(z), and \\bar B(z) are given beyond the trivial zero choice. Please define the stability notion and give at least one nontrivial class of target systems that satisfy the assumption.","section":"Theorem 6"},{"comment":"In the definition of \\bar a_0^-(z) in (16), the expression a_0^-(\\lambda_1^-/\\lambda_2^-\\,z,t) contains a time argument t even though the left-hand side depends only on z. This appears to be a typo and should be corrected.","section":"Section 3.3, Eq. (16)"},{"comment":"The phrase 'input-to-state stable' is used for the internal dynamics (45) without a definition adapted to the PDE setting. Please provide a definition or a reference to the relevant notion for infinite-dimensional systems.","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper presents an interesting and potentially valuable extension of backstepping control via dynamic extensions. The main issue is the incompleteness of the decoupling proof in Section 4.2; once the reduced kernel well-posedness and the L^{-+}=0 cancellation are established rigorously, the central claim would be sound. The PDE-ODE section also needs either full results or an explicit downgrade to a sketch. I would not reject, but the current version is not yet acceptable for publication in a journal of Automatica's standards."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look. The genuinely new piece is the dynamic extension that homogenizes all transport velocities of a heterodirectional hyperbolic system, and then uses a static feedback of the extended state to assign arbitrary in-domain couplings and to do input-output decoupling. That is more than a repackaging of Redaud et al.: it turns their constant-velocity result into a systematic construction for spatially varying velocities, and it makes the decoupling design look natural. The derivation in Section 3 is careful and the example is genuine, not a cartoon.\n\nThe stability theorem rests on standard backstepping plus the kernel well-posedness cited from Hu et al., so that part is solid. The decoupling section is the soft spot. The paper states that if L^-- and L^-+ solve (38a), (38b), (38d) with \\bar B=0, then (42) gives the decoupled form (43). It does not prove that this reduced kernel system is well-posed, and a first pass makes you worry about an extra \\bar A^+_0 contribution coming through the unchanged \\chi^+ dynamics. That worry does not survive: with \\bar B=0, the homogeneous PDE for L^-+ together with the diagonal boundary condition L^-+(z,z)=0 forces L^-+=0 on the whole triangle. So the extra term vanishes and the reduced kernel equations reduce to a simple well-posed problem for L^-- with boundary data at \\zeta=0. The authors should have said this in one sentence; the omission is minor but it is the only place where the reader has to fill in a real gap.\n\nRemark 7 on minimum-time preservation is asserted without proof. The claim is plausible and in fact true by the bounded invertibility of the Volterra transformation plus the zero target system, but again a sentence or two would settle it. Section 4.3 on PDE-ODE systems is more of a sketch than a theorem; fine as an outlook, but do not treat it as a proved transfer.\n\nWho gets value: people working in backstepping boundary control of hyperbolic PDEs, especially those interested in target-system flexibility and decoupling. The paper deserves a serious referee. It is a conditional accept, not a reject: fix the two missing justifications, tighten the PDE-ODE sketch, and it will be a useful contribution to Automatica or equivalent.","headline":"A solid, genuinely new backstepping design that uses dynamic extensions to homogenize transport velocities; the decoupling section is terser than it should be, but the apparent gap closes on inspection.","tokens_in":16774,"tokens_out":5035,"would_cite":true,"duration_ms":49103,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["93C20","93B52","93D15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A dynamic extension brings heterodirectional hyperbolic systems into a form where a static backstepping feedback assigns arbitrary target dynamics and complete input-output decoupling.","keywords":["dynamic extension","backstepping","heterodirectional hyperbolic systems","boundary control","input-output decoupling","transport velocity homogenization","PDE-ODE systems","state feedback design"],"falsifier":"Take a heterodirectional system with $n^-=n^+=2$, spatially varying speeds, $\\mathop{\\mathrm{rank}} Q_0=2$, and non-vanishing $\\bar{A}^-_0(z)$; solve the reduced kernel equations (38a), (38b), (38d) with $\\bar{B}(z)=0$ and check for a piecewise continuous $L^{--}$ and $L^{-+}$. If no such solution exists, the proposed decoupling transformation (42) is not well defined and the complete decoupling claim fails for that system.","tokens_in":15783,"feed_emoji":"🎛️","tokens_out":5861,"duration_ms":49405,"temperature":0.7,"pith_summary":"General heterodirectional hyperbolic systems—transport PDEs whose components travel in opposite directions with different speeds—are usually controlled by static state feedback, which forces the target closed-loop system to contain specific local terms and makes complete input-output decoupling impossible. This paper claims that by first adding a dynamic extension, i.e., a controller that is itself a set of transport equations, one can homogenize all transport velocities so that every component travels with the slowest speed. On the extended system, a static backstepping feedback of the extended state assigns a freely parametrized closed-loop dynamics with arbitrary in-domain couplings, and a decoupling feedback makes each output component track its own delayed reference. If the kernel equations used in the design are well posed, the result is a systematic method for dynamic state feedback design, including for coupled PDE-ODE systems, without losing minimum-time stabilization.","feed_headline":"A controller extension fully decouples hyperbolic boundary control systems","feed_subtitle":"A backstepping static feedback on the extended state assigns arbitrary target dynamics and preserves minimum-time convergence.","key_machinery":"The load-bearing object is the dynamic extension, a controller dynamics of transport equations with speeds chosen so that all components of the extended state share one left-going speed $\\lambda^-_{n^-}$ and one right-going speed $\\lambda^+_{n^+}$ on $[0,1]$. This homogenization makes the kernel equations for the backstepping transformation scalar-like: each block of the kernel $L(z,\\zeta)$ has only one characteristic family, so boundary values at $\\zeta=z$ and $\\zeta=0$ determine the solution. The Volterra integral transformation (35), with matrix gain $M(z)$ and kernel $L(z,\\zeta)$, then carries the extended system into the target system (34) with freely chosen matrices $B_0,B_1,B(z),\\bar{B}(z)$. The kernel equations (38) with boundary conditions determined by $M$ and $\\bar{A}_0$ are the equations whose solvability the whole construction rests on.","core_discovery":"The central claim is that dynamic extensions lift the well-known limitation of static backstepping for heterodirectional hyperbolic systems. A preliminary Volterra transformation removes in-domain coupling, and the dynamic extension is chosen so that all left-going components of the extended state propagate with the slowest left-going speed and all right-going components with the slowest right-going speed on the same unit interval. In the resulting dynamically extended system, a Volterra integral transformation with a matrix gain $M(z)$ and kernel $L(z,\\zeta)$ maps the extended system into a target system whose in-domain coupling matrix $\\bar{B}(z)$, boundary matrices $B_0,B_1$, and distributed boundary kernel $B(\\zeta)$ can be chosen freely. This is what enables assignment of general closed-loop dynamics and, as a special case, complete input-output decoupling with $y(t)=x^-(0,t)$ tracking $\\bar{v}(t-\\varphi^-_{n^-}(1))$ while the internal dynamics remains input-to-state stable. The same modular construction is claimed to transfer to PDE-ODE systems by combining the extension with an existing preliminary transformation. The paper's own condition is that the relevant kernel equations admit a solution; for the general target-system assignment this is proven under $\\mathop{\\mathrm{rank}} Q_0=n^-$, while the decoupling special case is asserted conditionally.","pith_inferences":["This suggests that dynamic extensions could be used to shape not only transport velocities but also the characteristic boundary maps, for instance to enforce passivity or prescribed relative-degree properties.","A natural testable extension is to compute the decoupling kernel explicitly for general $n^-=n^+=2$ systems with spatially varying speeds; the paper demonstrates the construction on one concrete example rather than proving solvability for all such systems.","The decoupling construction resembles a PDE analogue of the Byrnes-Isidori normal form; making that analogy precise could yield a systematic zero-dynamics assignment procedure for hyperbolic systems.","One could investigate whether the homogenization idea extends to systems with non-strictly ordered or coincident velocities, where the characteristic decomposition changes qualitatively."],"forward_implications":["Static state feedback of the original state cannot assign target systems with arbitrary in-domain couplings; with the dynamic extension, these couplings become free design parameters.","Complete input-output decoupling of $y(t)=x^-(0,t)$ with respect to a new input $\\bar{v}(t)$ is achieved, with closed-loop behavior $y(t)=\\bar{v}(t-\\varphi^-_{n^-}(1))$.","Minimum-time convergence is preserved: the homogenizing extension delays control action but does not increase the lower bound $T_{\\min}=\\varphi^-_{n^-}(1)+\\varphi^+_{n^+}(1)$.","The same design applies to hyperbolic PDE-ODE systems, so the ODE at the unactuated boundary can be stabilized while the PDE part enjoys the same decoupling or target-system freedom."],"supporting_citations":[{"why":"Supplies the kernel-equation framework and the observation that static multivariable backstepping forces local terms in the target system.","marker":"[11,10]"},{"why":"Shows that a time-affine transformation can assign arbitrary in-domain couplings, the phenomenon this paper reinterprets as a dynamic extension.","marker":"[21,22]"},{"why":"Defines the minimum control time for heterodirectional systems, the bound that the dynamic extension preserves.","marker":"[1]"},{"why":"Establishes finite-time stabilization of general linear hyperbolic balance laws, justifying the stability properties of the chosen target dynamics.","marker":"[3]"},{"why":"Provides the preliminary transformation that puts the PDE-ODE system into the cascaded form on which the dynamic extension is applied.","marker":"[7]"},{"why":"Gives existence and uniqueness of solutions to the matrix initial-value problem for $M(z)$.","marker":"[15]"},{"why":"Shows how to solve the scalar-like kernel equations by characteristics and successive approximations, the technique used for the homogenized kernel equations.","marker":"[24,4]"},{"why":"Motivates dynamic compensation for input-output decoupling in finite dimensions, the idea transferred to PDEs here.","marker":"[19]"}],"fun_headline_variants":["Dynamic extensions unlock full IO decoupling for hyperbolic control","Backstepping with dynamic extensions: full input-output decoupling","Hyperbolic control: dynamic extensions enable arbitrary target dynamics","Dynamic extension method decouples hyperbolic boundary systems","Backstepping dynamic extensions: complete IO decoupling for hyperbolic PDEs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The decoupling controller exists only if the reduced kernel equations that remove the local coupling in the left-going part have a solution; the paper assumes this without presenting a well-posedness proof for that special case.","fun_headline_variants_meta":{"raw":{"variants":["Dynamic extensions unlock full IO decoupling for hyperbolic control","Backstepping with dynamic extensions: full input-output decoupling","Hyperbolic control: dynamic extensions enable arbitrary target dynamics","Dynamic extension method decouples hyperbolic boundary systems","Backstepping dynamic extensions: complete IO decoupling for hyperbolic PDEs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000757,"raw_usage":{"total_tokens":3400,"prompt_tokens":1014,"completion_tokens":2386,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":630,"completion_tokens_details":{"reasoning_tokens":2305}},"tokens_in":630,"tokens_out":2386,"duration_ms":16177,"temperature":1.0,"reasoning_tokens":2305,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T10:41:09.632883+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a heterodirectional system with $n^-=n^+=2$, spatially varying speeds, $\\mathop{\\mathrm{rank}} Q_0=2$, and non-vanishing $\\bar{A}^-_0(z)$; solve the reduced kernel equations (38a), (38b), (38d) with $\\bar{B}(z)=0$ and check for a piecewise continuous $L^{--}$ and $L^{-+}$. If no such solution exists, the proposed decoupling transformation (42) is not well defined and the complete decoupling claim fails for that system.","supporting_citations":[{"cited_title":"Auriol and F","cited_arxiv_id":null,"evidence_quote":"Defines the minimum control time for heterodirectional systems, the bound that the dynamic extension preserves."},{"cited_title":"Coron, L","cited_arxiv_id":null,"evidence_quote":"Establishes finite-time stabilization of general linear hyperbolic balance laws, justifying the stability properties of the chosen target dynamics."},{"cited_title":"Deutscher, N","cited_arxiv_id":null,"evidence_quote":"Provides the preliminary transformation that puts the PDE-ODE system into the cascaded form on which the dynamic extension is applied."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives existence and uniqueness of solutions to the matrix initial-value problem for $M(z)$."},{"cited_title":"Morse and W.M","cited_arxiv_id":null,"evidence_quote":"Motivates dynamic compensation for input-output decoupling in finite dimensions, the idea transferred to PDEs here."}],"review_version":1}